I have been waiting for a long time for the companies running the vaccine trials to release data on what antibody level is protective. They took blood samples after the second shot, and have been monitoring people for breakthrough infections since then, so they should be able to go back to the prior blood samples and look at antibody levels for those who were infected in spite of being vaccinated.
Today we have that data from the AZ trial. They analyzed 171 breakthrough infections and 1404 controls without breakthrough infections, with respect to various types of antibody levels measured at 28 days post second shot.
Of course, a key factor here is that people's antibody levels decline pretty quickly from the day 28 peak, so this analysis is not based on antibody levels right before they had a breakthrough infection, but where they were at that early point. So if you got the AZ vaccine and want to do a comparison of where you would fall in this experiment, you'd need to know your level at day 28 post second shot. Because different vaccines have different rates of decay, these results probably will not hold true for other vaccines.
For the AZ trial in particular, antibody levels had decayed to 66% of their 28-day peak by day 90, and 36% of the peak by day 180. The study to find correlates of vaccine efficacy followed people for 95 days after the 28-day blood draw, which is about four months from the second shot.
The top-line result was that the threshold for 80% efficacy against symptomatic covid was 264 WHO international units, using standard 20-136. That translates to 203 U/ml on the Roche scale. The confidence interval was 83-620 U/ml.
Using the average decay rates above, 203 U/ml would decay to 134 U/ml by day 90. So 80% efficacy overall for a person who starts at 203 U/ml, then goes to 134 U/ml at three months is really not that bad.
People who had Pfizer or Moderna are averaging about 600 U/ml after six months.
They did give other efficacy levels based on other 28-day levels, which I have put into a graph above.
They did not find any threshold at which total immunity was achieved, which is not too surprising. There was also no correlation between antibodies and asymptomatic infection. The AZ vaccine seems to not be as good as the RNA vaccines at preventing asymptomatic infection, so that is also not too surprising.
Link to study:
https://www.medrxiv.org/content/10.1101/2021.06.21.21258528v1.full.pdf
Also, here is the study where I found the decay rate for the AZ vaccine:
https://www.thelancet.com/action/showPdf?pii=S0140-6736(21)00432-3